SR21  Redes Neurais, Algoritmos Genéticos e Lógicas Nebulosa/Neural
 
 Titulo:
NEURAL CRACK DETECTION FOR ACTIVELY CONTROLLED FLEXIBLE BEAM
 
Resumo :
ABSTRACT. FAULT DETECTION STUDIES OF MECHANICAL STRUCTURES HAS BEEN INTENSIFIED IN THE LAST TWO DECADES. METHODS BASED ON RESIDUAL GENERATION HAVE PRESENTED GOOD RESULTS, MAINLY WHEN ROBUSTNESS REQUIREMENTS ARE INCLUDED. AN ARCHITECTURE BASED ON OUTPUT OBSERVERS FOR RESIDUAL GENERATION IS PROPOSED FOR MONITORING OF A CONTROLLED FLEXIBLE BEAM, IN ORDER TO DETECT A CRACK. ADAPTIVE THRESHOLDS ARE IMPLEMENTED TO AVOID FALSE ALARMS, WITH VALUES CALCULATED EACH ITERATION, TRYING TO COMPENSATE FOR THE UNKNOWN DISTURBANCES THAT TEND TO ALTER THE REGULAR OPERATION OF THE PLANT. ON THE OTHER HAND, THE CRACK FORMULATION IMPLIES IN COMPLEX MODELS AND BRINGS HIGH COMPUTATIONAL COST THAT COULD MAKE THE WHOLE PROCESS SLOW. TO AVOID THIS PROBLEM, IT IS PROPOSED THE USE OF NEURAL OBSERVERS, EXPECTED TO GENERATE FASTER AND MORE ACCURATE FAILURE DETECTION. FURTHERMORE, OBSERVER-BASED RESULTS ARE ALSO PRESENTED, TO COMPARE WITH THE NEURAL RESIDUAL GENERATION. IT IS POSSIBLE TO CONCLUDE THAT NEURAL OBSERVER IS AN EFFICIENT METHOD FOR CRACK DETECTION IN FLEXIBLE BEAMS. KEYWORDS: FAULT DETECTION, ADAPTIVE CONTROL, NEURAL OBSERVER.  
 
Autores :
Alves Jr, Marco Antonio Oliveira
Nobrega, Euripedes Oliveira
 
 
Trabalho Completo :

 

COBEM99 | Comissão  Organizadora | Palestras | Sessões Técnicas | Autor | Simpósios e Sessões Especiais | Revisores
Título  dos  Trabalhos | Local